On the development of statistical models for assessing projects, portfolios and dissertations

Numerous assessment formats have evolved in higher education in recent years – many inspired by task-related activities in the workplace. Some are not new: at Masters level, the dissertation is long-established, whereas at undergraduate level, the use of projects and portfolios is becoming increasingly fashionable. Implementing these different forms of assessment is not always easy however - even when strict rubrics are enforced. As a consequence, double-marking is frequently used to offset the subjectivity of marks awarded. Unfortunately, this strategy too is not without its difficulties – as recent studies have shown – especially when there is fundamental disagreement between first and second examiners. Focussing on this issue of inter-marker conflict, a series of simple statistical models are developed to help assess how final marks might be more objectively determined.